Economic Ideology and the Mediating Role of Electoral Institutions

We draw on existing research on domestic responses to economic globalization to derive clear predictions regarding the ideological shifts of mainstream legislators representing regions hard hit by import shocks. Given that politicians from mainstream parties focus primarily on economic issues and building on work showing legislators respond to trade influxes along the economic dimension (Autor et al. Reference Autor2020; Campello and Urdinez Reference Campello and Urdinez2021; Feigenbaum and Hall Reference Feigenbaum and Hall2015), we expect that, all else being equal, non-extremist political elites' responses to trade shocks will be reflected primarily in economic ideological shifts. There are two economic policy mechanisms available to legislators to address the dislocations created by import shocks: legislators can advocate expanding compensation and/or greater protectionism. While the literature has increasingly argued that both voters (Colantone and Stanig Reference Colantone and Stanig2018) and (non-European) legislators (Campello and Urdinez Reference Campello and Urdinez2021; Feigenbaum and Hall Reference Feigenbaum and Hall2015) are likely to favour protectionism in response to trade shocks, in the European context, legislators' economic policy responses are more limited. For all European Union (EU) member states, decisions about trade policy are made at the EU level, with the EU Commissioner for Trade acting on behalf of all member states (Nanou and Dorussen Reference Nanou and Dorussen2013). Legislators in EU member states thus have limited influence over trade policy and cannot credibly promise to introduce protectionist measures since voters in these countries are aware politicians' hands are tied on these issues (Hellwig Reference Hellwig2014; Meyerrose Reference Meyerrose2020, Reference Meyerrose2023). Therefore, we expect them to favour compensation- and redistribution-oriented policy solutions to the dislocations created by import shocks.

Localized import penetration increases economic and employment volatility, which in turn makes voters more likely to demand increased government spending to compensate for these economic dislocations (Rodrik Reference Rodrik1998; Ruggie Reference Ruggie1982). These demands for compensation make voters more likely to support left-leaning politicians and parties that advocate welfare state expansion and redistribution. Indeed, there is evidence that the economic insecurities created by import shocks have driven affected voters to favour compensatory policies associated with traditional parties on the left (Rommel and Walter Reference Rommel and Walter2018; Scheve and Serlin Reference Scheve and Serlin2023) and left-wing parties and candidates that universally reject neoliberalism, consumerism, and globalized capitalism have become increasingly electorally successful (Bale and Dunphy Reference Bale and Dunphy2011; Hopkin Reference Hopkin2020). This, in turn, has led to greater government spending (Dreher, Sturm, and Ursprung Reference Dreher, Sturm and Ursprung2008; Garrett Reference Garrett1998; Rodrik Reference Rodrik1998) and an expansion of redistributive policies (Bergh and Nilsson Reference Bergh and Nilsson2010; Leibrecht, Klien, and Onaran Reference Leibrecht, Klien and Onaran2011; Meinhard and Potrafke Reference Meinhard and Potrafke2012).

While import shocks directly impact employment outcomes for only a subset of workers within affected regions, the effects of these shocks on voters' economic preferences extend beyond the individual level. Localized trade shocks can trigger sociotropic considerations that influence not just the economic policy preferences of impacted workers but also the preference of all voters in these regions who are concerned with the economic well-being of their community (Rickard Reference Rickard2022). Indeed, there is evidence that awareness of rising unemployment leads many voters in a region, regardless of their employment status, to hold the government accountable for their locality's shifting economic landscape (Mansfield and Mutz Reference Mansfield and Mutz2009). This, in turn, creates incentives for legislators representing these areas to shift their economic ideological position. Therefore, our baseline prediction is:

Where institutions increase incentives for elected officials to be responsive to local demands, the effect of import shocks on legislator ideology should be magnified. Research shows one type of institution in particular – the electoral system – creates distinct incentives for how politicians respond to local economic conditions (Breunig, Grossman, and Hänni Reference Breunig, Grossman and Hänni2022; Chang et al. Reference Chang2010; Katz Reference Katz1997; Shugart and Taagepera Reference Shugart and Taagepera1989; Taagepera Reference Taagepera1973). Since Duverger (Reference Duverger1954), political scientists have highlighted how electoral systems can induce changes in elite ideology through their mechanical effects on vote-seat elasticities. In majoritarian systems, small vote swings can dramatically distort the relationship between votes and seat shares (Rogowski and Kayser Reference Rogowski and Kayser2002). Because identical vote distributions are often translated into different seat allocations, depending on the electoral system, in majoritarian systems (where small vote swings can dramatically distort the relationship between vote and seat shares), alienating voters by failing to offer policies that address the negative effects of import shocks entails a higher risk for politicians (Wlezien and Soroka Reference Wlezien and Soroka2012).

Existing work suggests another channel through which majoritarianism promotes greater responsiveness: the fact that candidates in majoritarian districts can campaign, at least to some extent, on personal characteristics and are thus less dependent on their party for re-election (Carey Reference Carey2007; Carey and Shugart Reference Carey and Shugart1995; Katz Reference Katz1997; Mitchell Reference Mitchell2000). In districts where the incumbency advantage plays a more pivotal role (Hainmueller and Kern Reference Hainmueller and Kern2008), we similarly expect legislators to face greater incentives to respond to trade-generated dislocations, irrespective of the position taken by their political party (Sieberer Reference Sieberer2010).

Although majoritarian electoral systems create incentives for greater elite economic ideological shifts in regions hard hit by imports, the literature also suggests the extent to which these electoral incentives influence politicians' strategies is contingent on the competitiveness of the election. According to the ‘marginality hypothesis’, legislators elected by narrower margins (for example, those in more competitive districts) will be more attentive to voter demands due to the possibility of voter backlash. Indeed, focusing on legislative responses to trade, Feigenbaum and Hall (Reference Feigenbaum and Hall2015) find that US legislators in competitive electoral districts shift their ideology. Drawing on this research, we predict the following:

While the literature offers clear expectations of how mainstream legislators will respond economically to trade shocks, the extent to which these legislators shift their ideological position with respect to the second cultural dimension is contingent on the political competition they face. We argue that mainstream legislators will be more likely to adjust their cultural positions in response to trade influxes when they face electoral competition from extremist competitors who, as noted above, define themselves primarily with respect to cultural issues rather than economic ones. Absent competitors that make this second dimension salient, legislators will prefer to focus on the economic policy issues central to their political identity.

Incorporating Cultural Ideology: The Role of Extremist Challengers

In representative democracies, elections create incentives for politicians to adjust their ideological positions. However, the nature and context of these elections also influence these incentives. In addition to electoral institutions and the competitiveness of a constituency, the type of competition politicians face should also shape responses to rising import competition (Watson Reference Watson2015). As extremist parties grow in popular support and, as a result, increase the salience of cultural issues, they pose a direct electoral challenge to mainstream politicians. Indeed, research finds that, when threatened electorally, traditional parties may opt for an accommodative approach wherein they shift their cultural ideology to more closely match that of their extremist competitors in the hopes of drawing in or winning back voters (Meguid Reference Meguid2005). In a direct test of these dynamics, Abou-Chadi and Krause (Reference Abou-Chadi and Krause2020) find mainstream parties are more likely to emphasize anti-immigration positions when the populist radical right received higher vote shares in the previous election.

While work on mainstream party responses to extremists focuses primarily on the far right, the rise of the far left in Europe in recent years suggests a strong radical left presence might similarly influence mainstream elites' policy positions. Building on previous findings to consider individual legislators (rather than parties), we expect mainstream legislators will adopt accommodative strategies when confronted with electoral threats from the far right and the far left. However, the types of threats posed by radical left and right competitors are somewhat distinct.

Populist radical right politicians primarily emphasize second-dimension cultural issues such as immigration. Their stances resonate with working-class voters experiencing significant economic hardship because of globalization, particularly in areas hit hardest by trade (Dippel, Gold, and Heblich Reference Dippel, Gold and Heblich2015; Malgouyres Reference Malgouyres2014). Given the electoral salience of immigration, politicians across the political spectrum have increasingly advocated anti-immigrant policies to win votes and compete with the far right (Abou-Chadi Reference Abou-Chadi2016; Abou-Chadi and Krause Reference Abou-Chadi and Krause2020; Meguid Reference Meguid2005; Schain Reference Schain, Schain, Zolberg and Hossay2002; Spoon and Klüver Reference Spoon and Klüver2020; van Spanje Reference van Spanje2010). Therefore, we predict the following:

While the far right defines itself primarily with respect to cultural issues, the far left is more ideologically complex. In Europe, the far left is comprised of both post-materialist niche parties, such as the Greens, who focus primarily on second-dimension issues, and materialist radical-left parties, such as the Communists, who define themselves predominantly with respect to economic topics. Despite having distinct ideological foundations, both far-left parties compete for left-wing voters who have become disenchanted with the mainstream left. In times of economic prosperity, the Greens emphasize their core environmental issues. However, during times of economic hardship – such as those that result from import shocks – Communists are particularly successful in attracting voters sympathetic to their economic platforms. In these contexts of economic decline, Green parties shift to emphasizing economic issues to compete with the materialist-oriented far left and attract additional left-leaning voters (Spoon and Williams Reference Spoon and Williams2021). Therefore, while we expect far-right competitors to impact a mainstream legislator's cultural ideological position, we anticipate individual politicians from mainstream parties will shift to the left on both the economic and cultural dimensions when confronted with a strong, far-left challenger.Footnote ⁵ Formally:

Finally, we also evaluate the possibility of differential responses to competitive contexts across partisan identities. Changing ideologies is politically risky for any politician, as it can signal a weak commitment to core principles. Mainstream politicians should only accommodate these extremist positions if they face real and immediate electoral threats; they should be particularly likely to accommodate policy positions closer to their core ideological foundations (Han Reference Han2015). Therefore, we expect leftist politicians will be more responsive to a far-left challenger than their mainstream right-wing counterparts, while right-leaning politicians will be more vulnerable to challenges from the far right and respond accordingly. In other words, we predict:

To summarize, we predict that, on their own, import shocks will make mainstream legislators more likely to shift their economic ideology to the left, reflecting considerations surrounding redistribution and compensation. However, the competitive context in which these legislators operate should also impact the relationship between import shocks and ideology. First, since majoritarian electoral institutions create incentives for greater responsiveness than PR systems, we expect that legislators elected under majoritarian rules will be particularly likely to shift their economic ideology to the left; this should especially be the case in competitive majoritarian systems. Second, we predict mainstream legislators' ideological shifts in response to import shocks will be influenced by competition from extremist competitors. Specifically, we expect competition from the far right will pull legislators in trade-affected regions to the right on the cultural dimensions, while competition from the far left will induce further shifts to the left on both economic and cultural issues. Right-wing politicians will be particularly responsive to far-right challenges, while left-wing candidates will be more susceptible to challenges from the far left. These predictions are summarized in Table 1.

Table 1.

Summary of Predictions

In testing these hypothesized relationships between import shocks, local competitive context, and legislator ideology, it is important to consider the potential mechanisms driving any observed ideological shifts. On the one hand, movements in legislative ideology could be driven by the election of new legislators that are further to the left/right than their predecessor (‘replacement’). Indeed, a substantial literature in political economy finds that economic downturns disadvantage incumbent politicians (Brooks and Brady Reference Brooks and Brady1999; Lewis-Beck and Paldam Reference Lewis-Beck and Paldam2000), as voters punish incumbents at the polls for the adverse economic outcomes, including those caused by greater trade exposure (Jensen, Quinn, and Weymouth Reference Jensen, Quinn and Weymouth2017; Margalit Reference Margalit2011; Reference Margalit2013). There could also be a selection mechanism at work, as competition for votes induces incumbent politicians in trade-exposed districts to strategically shift their ideological stances in response to changing voter preferences (‘realignment’). Below, we test the extent to which our results are driven by realignment or replacement dynamics.

Dependent Variable: Ideology of French Legislators

We use a spatial model to capture ideological shifts among French senators. Applications of spatial models are extensive in the US Congress (Poole and Rosenthal Reference Poole and Rosenthal1985) but have also been applied in legislatures with more than two political parties (Hix and Noury Reference Hix and Noury2009; Meyerrose Reference Meyerrose2018).Footnote ¹¹ However, only a subset of these models allows for the intertemporal comparisons of ideology needed to test our argument regarding shifts in legislative positions over time in response to import shocks. We use a Bayesian dynamic item response (IRT) model to measure senators' ideological changes. The dynamic IRT model, developed by Martin and Quinn (Reference Martin and Quinn2002), is a more flexible alternative to DW-NOMINATE, another intertemporal spatial model (Poole and Rosenthal Reference Poole and Rosenthal2001). The dynamic IRT model allows individuals' ideologies to change non-monotonically across legislative sessions (Clinton, Jackman, and Rivers Reference Clinton, Jackman and Rivers2004) and produces more precise point estimates than DW-NOMINATE.

Like American senators, not all French senators are elected simultaneously. American politics scholars count each two-year congressional session as a legislative session for the US Senate; we follow a similar strategy. Because French parliamentary sessions only last nine months, we divide the French Senate into sessions coinciding with periods between elections (three years).Footnote ¹² We assign a session ID to each three-year period between the Senate elections during the Fifth Republic. Our sample includes voting data from seven sessions: 13 (1996–1998) through 19 (2014–2017).

To estimate the dynamic IRT model, we scraped and cleaned information on public roll call votes for all but two of the Senate sessions between 1996 and 2017.Footnote ¹³ The result is a dataset of 3,589 roll call votes with information on how each senator voted on a given bill and a text-based description of each bill's content.

We predict mainstream legislators representing import-exposed departments will be more likely to shift their economic ideological position to the left and, depending on their local context and partisan attachment, their cultural position to either the left or the right.Footnote ¹⁴ We estimate legislators’ ideological positions on a subset of relevant roll call votes to ensure we capture shifts in the economic and cultural dimensions. Drawing on the economic and cultural categories identified in the Comparative Manifesto Project (Volkens et al. Reference Volkens2020), we use the text descriptions to hand code each of the 3,589 bills as either economic, cultural, or neither.Footnote ¹⁵ The result is two mutually exclusive sets of roll call votes: 2,227 economic and 613 cultural.

We validate our hand coding using the Comparative Agendas Project data (Baumgartner et al. Reference Baumgartner, Breunig and Grossman2019), which codes a subset of the bills in our dataset by topic.Footnote ¹⁶ Since we identify relatively few cultural bills, we adopt a model-based approach to validate our codings further. DW-NOMINATE models allow the researcher to define the number of ideological dimensions ex-ante and then use patterns in the data to place legislators along the specified number of dimensions. We estimate a DW-NOMINATE model along two dimensions for the French Senate, finding that the first dimension alone correctly classifies about 95 percent of all roll call votes in our dataset. Although somewhat surprising, this gives us further confidence that our codings, which are heavily skewed toward economic bills, are accurate.

To estimate ideological positions in the French Senate, we apply an approximation of the dynamic IRT model using the emIRT package in R (Imai, Lo, and Olmsted Reference Imai, Lo and Olmsted2016).Footnote ¹⁷ In preparing the data, we follow conventional practices and exclude individual roll call votes for which the vote was either 97.5 percent for or against. We also remove legislators who did not vote at least twenty-five times within a given parliamentary session.Footnote ¹⁸ Finally, given the French Senate's strict legislative voting rules, our view is that abstentions essentially equate to ‘nay’ votes in this context and we code them as such.Footnote ¹⁹

The dynamic IRT model requires the modeler to specify a set of prior distributions. We set priors based on a legislator's party affiliation to orient the ideological space. These priors are only used for the first session a senator serves, so they are not overly restrictive. We estimate two IRT models: one for the economic bills and another for the cultural ones.Footnote ²⁰ We operationalize our dependent variables, economic and cultural ideological shifts, simply as changes in the relevant IRT scores, both at the department and individual senator levels.Footnote ²¹

Independent Variables

To measure trade-induced labour market shocks, we use the shift-share Imports per Worker (IPW) measure pioneered by Autor, Dorn, and Hanson (Reference Autor, Dorn and Hanson2013) and Autor et al. (Reference Autor2020) to estimate local-level exposure to imports from China. The intuitive idea behind this approach is that local labour markets are differentially affected by the growth in imports from low-wage countries depending on their prior industry specialization. Any changes in imports at the industry level will disproportionately impact areas with higher degrees of employment in that industry.

As Fig. 2 shows, there is a substantial geographic variation with respect to trade exposure in France. Here, we see import competition in France operates on roughly a southwestern to northeastern axis. Departments in the northeast have been particularly hard hit. In contrast, the southwestern regions, less intensive in import-competing industries, have far lower IPWs.

Figure 2.

Change in Trade Exposure, 1996–2017. This map depicts changes in imports per worker from China in each of France's 95 metropolitan departments disaggregated across 38 employment sectors (NA A38). Source: COMEXT and INSEE.

Following previous studies, we address the possible endogeneity of legislative voting patterns to the import shock by instrumenting IPW using the growth in imports from China to five other wealthy European countries: Germany, Denmark, Sweden, Italy, and the UK (Dauth, Findeisen, and Suedekum Reference Dauth, Findeisen and Suedekum2014).Footnote ²²

Our other main independent variables of interest relate to local context: electoral institutions, electoral competitiveness, and patterns of party competition. We use French Senate election data to construct measures indicating whether a department held elections under majoritarian (as opposed to PR) electoral rules, as well as constituency-level vote margins and radical party vote share measures from the prior election.Footnote ²³ In addition to our main independent variables, and as shown in Table 2, our models include a measure of the start-of-period political tendency of the department, as well as controls for demographic and economic factors, including a department's start-of-period logged total population, percent female, age structure, percent immigrant population, and percent employed in industry. This data was obtained from INSEE's Estimations de Populations. We also estimate models with two sets of political controls. The first includes a measure of department-level competitiveness, department magnitude, degree of local office holding, and an indicator for a majoritarian electoral system. We also include models that control for other changing features of the political context that might affect legislative behaviour, including indicators for whether, in a particular session, a department experienced a change in its electoral system, there was an expansion of seats, senators in a department were up for re-election in the subsequent session, and if the legislature was divided.

Table 2.

Models and Variables

Finally, for all models, we include two-way unit and time-fixed effects,Footnote ²⁴ and we report heteroskedasticity-robust standard errors clustered at the region level to account for spatial autocorrelation. In Appendix L, we report full regression tables, including F-statistics based on robust to weak instrument tests (Pflueger and Wang Reference Pflueger and Wang2015).

Estimation Strategy

To estimate the relationship between trade exposure and legislative behaviour, we follow standard practice by adopting a first-differences model, which allows us to control for time-invariant heterogeneity:

(1)$$\Delta {\rm IR}{\rm T}_{it} = \beta \Delta {\rm IP}{\rm W}_{it} + {X}^{\prime}_{it} + \epsilon _{it}$$

where it refers to department i from time t to t + 1. We defined ΔIPW_it previously. ΔIRT_it refers to the change in the department ideology score from time (session) t to t + 1. X_it indicates a series of departmental start-of-period political and demographic controls, while ɛ_it is an idiosyncratic shock we assume is uncorrelated with the regressors. We examine changes in import exposure and its effect on legislator ideology across all elections in our sample. We generate session-by-session IPW measures, IRT scores, and start-of-period controls for each department. This gives us five time periods: the change in trade exposure and legislative ideology from Senate sessions 13 to 14, sessions 14 to 15 and so on, up through session 19. We then pool the data and use stacked first-difference estimation to examine the effects of short-term shifts in localized trade exposure on changes in legislator ideology.Footnote ²⁵

The Mediating Effect of Electoral Institutions

As discussed in reference to Hypothesis 2, assuming reasonable levels of electoral competition, we expect the economic ideological effects of trade to be magnified in departments where senators are elected using majoritarian rules. To test this prediction, we report estimates from two sets of models. The first focuses on within-senator responses to import shifts, while the second reports department-level effects. Coefficients for the senator-level models are shown in Figs 4a and 4b. Figure 4a suggests that, overall, majoritarian electoral institutions have little effect on changes in economic ideology above and beyond the effects of trade. Figure 4b explores whether the effects of majoritarianism are contingent on the degree of electoral competition by reporting coefficients for the interaction of ΔIPW, majoritarianism, and department vote margin at different levels of electoral competitiveness. For the within-senator models, as political competition tightens in majoritarian districts, the senator's economic ideology shifts to the left in response to higher levels of import competition, providing support for Hypothesis 2.

Figure 4.

Imports, Electoral Systems and Competitiveness: Senator vs Department-Level Models.

Note: For Figs 4a and 4c, the plots report the coefficient on ΔIPW interacted with majoritarianism. Model 1 (top) includes economic and demographic controls (as defined in Table 2); Model 2 (middle) adds political controls A; and Model 3 adds Political controls B. Figs 4b and 4d report coefficients on the interaction of ΔIPW, majoritarianism and the vote margin of a department, at different levels of vote margin, using the full set of controls (i.e., Model 3). All models include unit and session fixed effects.

The models reported in Figs 4a and 4b analyze the response of sitting senators but do not consider how the election of new senators shapes ideological shifts within departments. We investigate these dynamics with department-level models, reported in Figs 4c and 4d. Figure 4c shows that as import penetration increases in majoritarian departments, the average economic ideology score of the department shifts further to the left compared to PR departments. Figure 5 shows the marginal effects of the fully saturated model. Interestingly, however, Fig. 4d shows this leftward shift takes place irrespective of the electoral competitiveness of the department. The finding that majoritarianism induces a leftward response, even under low levels of electoral competition, suggests the effects of electoral systems in this context may work not simply through what Duverger (Reference Duverger1954) termed the ‘mechanical’ channel of vote-seat shares but also through a ‘psychological’ channel involving strategic coordination by parties and/or voters over candidate selection. We briefly explore these dynamics in Appendix E and return to this possibility in the conclusion.

Figure 5.

Imports and Electoral Systems: Marginal Effects for Department-Level Models.

Note: Marginal effects for change in IPW × Majoritarianism using the full set of controls (economic + demographic controls, plus political controls A and B).

How Radical Party Competition Matters

Finally, in Hypotheses 3 and 4, we predict that mainstream legislators will adopt accommodative strategies when confronted with electoral threats from extremist competitors from both the far right and the far left. Figure 6 begins to explore these dynamics, reporting coefficients on models interacting ΔIPW with the electoral strength of radical left and radical right parties in the previous Senate elections in each department.Footnote ³⁰

Figure 6.

Imports and Radical Party Strength: Overall Results.

Results shown in Fig. 6a suggest that in departments with a pre-existing strong radical left, increases in IPW produce an overall modest leftward shift in economic ideology and a small leftward shift in cultural ideology. By contrast, Fig. 6b suggests that as local political support for the radical right increases, higher levels of import exposure shift the ideology of sitting senators to the right. This effect is particularly strong on the economic dimension but also notable on the cultural one.

Since dramatically changing political ideology is risky for any politician, we also predicted in Hypotheses 5 and 6 that mainstream senators on the right will be more responsive to challengers from the far right, while mainstream left senators will be more likely to accommodate competitors from the far left. To further explore these heterogenous treatment effects by party, in Fig. 7, we re-estimate the models from Fig. 6, but this time, we break out responses by party family with several noteworthy findings. First, Figs 7a and 7b find support for Hypothesis 6: mainstream senators on the left representing departments with a higher radical left vote share adopt accommodative strategies, shifting further to the left on economic and cultural dimensions.Footnote ³¹ Similarly, as predicted in Hypothesis 5, in Figs 7c and 7d, we find far-right competitors only significantly impact the economic and cultural ideology of right-leaning mainstream senators, while the results for left-leaning politicians are statistically insignificant.

Figure 7.

Imports and Radical Party Strength By Party Family.

We note the results in Figs 6b and 7c were unexpected. As outlined in Table 1, we did not predict the far right, which defines itself primarily with respect to second-dimension issues, to affect the economic ideology of any mainstream senators. Since this effect is driven by right-leaning mainstream senators, it is perhaps unsurprising that these senators would be willing to shift their economic and cultural positions to compete with the far right since this would not constitute a large change from their core positions. However, the direction of the shift is also puzzling since, as noted in Section ‘Import Shocks, Local Electoral Context, and Legislator Ideology’, the National Rally and other far-right parties in Western Europe have, over time, become associated with left-leaning economic platforms that emphasize welfare chauvinism. However, this shift to the economic left is a relatively recent phenomenon, precipitated by the fallout from the 2008 financial crisis (Mudde Reference Mudde2016). Prior to the 2010s, and thus during much of the time period covered in our analyses, most scholars argue that far-right parties in Western Europe adopted economically right-wing and neoliberal platforms (Kitschelt and McGann Reference Kitschelt and McGann1995; Kriesi et al. Reference Kriesi2012; Minkenberg Reference Minkenberg2000) that more closely align with the economic ideology of the mainstream right.